Sputum Gene Expression Reveals Dysregulation of Mast Cells and Basophils in Eosinophilic COPD.

Purpose: The clinical and inflammatory associations of mast cells (MCs) and basophils in chronic obstructive pulmonary disease (COPD) are poorly understood. We previously developed and validated a qPCR-based MC/basophil gene signature in asthma to measure these cells in sputum samples. Here, we measured this gene signature in a COPD and control population to explore the relationship of sputum MCs/basophils to inflammatory and COPD clinical characteristics. Patients and
Methods: MC/basophil signature genes (TPSAB1/TPSB2, CPA3, ENO2, GATA2, KIT, GPR56, HDC, SOCS2) were measured by qPCR in sputum from a COPD (n=96) and a non-respiratory control (n=17) population. Comparative analyses of gene expression between the COPD and the control population, and between eosinophilic COPD and non-eosinophilic COPD were tested. Logistic regression analysis and Spearman correlation were used to determine relationships of sputum MC/basophil genes to inflammatory (sputum eosinophil proportions, blood eosinophils) and clinical (age, body mass index, quality of life, lung function, past year exacerbations) characteristics of COPD.
Results: MC/basophil genes were increased in COPD versus control participants (CPA3, KIT, GATA2, HDC) and between eosinophilic-COPD and non-eosinophilic COPD (TPSB2, CPA3, HDC, SOCS2). We found all MC/basophil genes were positively intercorrelated. In COPD, MC/basophil genes were associated with eosinophilic airway inflammation (GATA2, TPSB2, CPA3, GPR56, HDC, SOCS2), blood eosinophilia (all genes) and decreased lung function (KIT, GATA2, GPR56, HDC).
Conclusion: We demonstrate associations of MCs and basophils with eosinophilic inflammation and lower lung function in COPD. These findings are consistent with prior results in asthma and may represent a new tool for endotyping eosinophilic-COPD.

Plain Language Summary

Mast cells (MCs) and basophils are important granulocytes that exert numerous immune functions, the dysregulation of which promote airway inflammation, bronchoconstriction and airway remodelling. The scarcity of these immune cells in the airways and the absence of MCs from circulation make them challenging to study in the clinical context and their inflammatory and clinical associations in COPD remain poorly understood. We have previously developed and validated a qPCR-based MC/basophil gene signature in asthma that reflects the abundance of these cells in sputum samples. Here, we apply this gene signature in a COPD cohort (n=96) and non-respiratory disease controls (n=17), demonstrating dysregulation of sputum MCs and basophils in COPD. Sputum MC/basophil-related gene expression was related to airway and systemic eosinophilic inflammation and lower lung function in COPD. These novel findings demonstrate potentially automatable PCR-based measure of airway MCs/basophils in COPD and reveal dysregulation of lumen MCs and basophils and associations with important clinical and inflammatory characteristics in COPD.

Introduction

COPD is the third leading cause of death and fifth cause of disability worldwide.1 Disease progression is marked by lung function decline and increased exacerbations of worsening symptoms.2 Inflammation plays an integral pathophysiological role in COPD. Patients can present with varied inflammatory patterns, the most well described being neutrophilic airway inflammation.3 However, eosinophilic inflammation, in sputum and blood, is also present in a significant number of individuals with COPD.4–8 Eosinophilic COPD (E-COPD) is associated with increased acute lung attacks,8,9 as well as higher healthcare utilization and costs.10 An understanding of the inflammatory pathways and mechanisms of E-COPD, and biomarkers that can support treatment and management is needed. Mast cells (MCs) and basophils exert numerous immunological functions associated with allergy, eosinophilic inflammation11,12 and airway remodelling.13 Recently, sputum MCs have been associated with asthma control and lung function14 and have been found to be dysregulated in severe asthma.15–17 Bronchial MC numbers are increased in smokers compared to non-smokers, with MC density correlating with airway remodelling18 and lung function.19 Sputum MC numbers are increased in current COPD smokers compared to ex-smokers20 and biopsy basophil numbers are increased in COPD patients compared to controls.21 However, the relationship of MCs and basophils to inflammation and clinical features of COPD, particularly E-COPD, is less understood. Further, the scarcity of MCs and basophils in airway-derived clinical samples such as sputum and the absence of mature MCs in circulation can make direct measurement difficult.

MCs and basophils possess distinctive and overlapping transcriptomic signatures expressed at greater levels compared to other immune cells in the lungs.22,23 We recently developed and validated a sputum MC/basophil gene signature, consisting of eight genes relating to varied MC/basophil immune functions (TPSAB1/TPSB2, CPA3, ENO2, GATA2, KIT, GPR56, HDC, SOCS2). qPCR-based measures of MC/basophil-related genes were directly related to MC and basophil abundance in sputum samples.17 Jiang et al further demonstrated a similar MC gene signature was directly related to biopsy MC abundance.24 These findings demonstrate the value of unique MC/basophil-related gene signatures to measure their abundance in complex primary airway samples such as sputum and bronchial biopsies.

Here we measured the validated sputum MC/basophil gene signature in COPD (n=96) and control populations (n=17) to explore the relationship of MCs and basophils to airway and systemic inflammation, clinical characteristics and exacerbation risk in COPD. We hypothesized that MC/basophil-related genes would be differentially expressed between control participants, eosinophilic and non-eosinophilic COPD and would be associated with systemic eosinophilia, clinical characteristics, and risk of lung attacks in COPD.

Methods

Study Design

In this observational, cross-sectional study, COPD participants were recruited via the respiratory ambulatory care clinics at John Hunter Hospital (NSW, Australia), the clinical research database of the Department of Respiratory and Sleep Medicine, John Hunter Hospital and by advertisement. Control participants were recruited by advertisement. Studies were approved by the Hunter New England Health Human Research Ethics Committee for COPD (12/12/12/3.06) and control (8/08/20/3.10) populations and research conducted complied with the Declaration of Helsinki. All participants provided written informed consent.

Participants

Participants required a doctor-confirmed diagnosis of COPD with documented evidence of incompletely reversible airflow limitation, and stable COPD at study visit. For detailed inclusion/exclusion criteria for COPD participants, see supplementary appendix S1. Control participants had no diagnosis of respiratory disease and had no overt respiratory infection at study visit. Current smokers were excluded from the study (COPD and controls).

Clinical Methods

Participants underwent previously published clinical methods,25 including sputum sample collection, blood collection for full blood count (peripheral blood eosinophil [PBE] count), spirometry (forced expiratory volume in one second [FEV1%], forced vital capacity [FVC%] and FEV1/FVC%), health status assessment (COPD Assessment Test [CAT] and St George’s Respiratory Questionnaire [SGRQ]) and past year exacerbation history (see for exacerbation definitions).

Sputum Processing and Inflammatory Phenotyping

Sputum was induced and processed as described previously26 (see supplementary appendix S1 for details). Airway inflammatory phenotypes were determined via differential cell counts as follows: neutrophilic inflammation (N-COPD) (≥61% neutrophils, <3% eosinophils); eosinophilic inflammation (E-COPD) (<61% neutrophils, ≥3% eosinophils); mixed granulocytic inflammation (MG-COPD) (≥61% neutrophils, ≥3% eosinophils); paucigranulocytic inflammation (PG-COPD) (<61% neutrophils, <3% eosinophils).27 Dichotomous categorisations include eosinophilic (E-COPD and MG-COPD) vs non-eosinophilic (NE-COPD, which combines N-COPD and PG-COPD). PBE categories were determined using a PBE threshold (PBE high ≥300 cells/µL vs low <300 cells/µL).

Laboratory Methods

Sputum qPCR-based gene expression methods have been previously described.28–30 Target gene expression normalized to the B-actin housekeeping gene. ΔCt (change in cycle threshold between target gene and B-actin) and 2−ΔCt (relative mRNA abundance) values were used in the analysis. For detailed laboratory methods see supplementary appendix S1.

Statistical Analyses

Two-group comparisons were analyzed using Student’s t-test (parametric) or Wilcoxon rank sum (non-parametric). Kruskal–Wallis with Bonferroni post-hoc correction was used for multiple groups. Categorical data were analyzed using Fisher’s exact test, with Fisher’s p-value reported when expected counts were <5. Spearman correlation coefficients were used. Relative mRNA abundance values (2−ΔCt) were used for comparisons between groups and correlation analysis. Gene expression values (ΔCt) were used for regression analysis due to approximate normal distribution. MC/basophil genes were analyzed individually and in a combinatorial gene metric (see supplementary appendix S1).31 Relationship of MC/basophil genes to inflammatory phenotypes was determined with univariate logistic regression and receiver operating characteristic (ROC) curve analysis. Least absolute shrinkage and selection operator (LASSO) multivariate regression was used to select important genes in a model (R Studio v.3.6.2). Participants were categorized by PBE threshold (PBE high ≥300 cells/µL vs low <300 cells/µL) and frequent (≥2) and infrequent (0–1) for total and severe exacerbations in prior 12 months. A p-value <0.05 was considered statistically significant. Data were analyzed using STATA v.15.1 and GraphPad Prism v7. The data that support the findings of this study are available from the authorship team upon reasonable request.

Results

Clinical Characteristics of COPD and Control Participants

Participant demographics for both groups are shown in Table 1. COPD individuals were older, with more ex-smokers who had a higher pack-year history (Table 1). Most COPD participants had severe disease (GOLD quadrant D, n=70 [72.9%]). Sputum neutrophil and eosinophil proportions and PBE counts were higher in COPD participants compared to controls.

Table 1

Clinical Characteristics of Study Sample Populations

	Controls	COPD	p-value
Sample number (n)	17	96
Clinical characteristics
Sex, female n (%)	7 (41.2)	35 (36.5)	0.71
Age, years (range)	39.5 (22.5–72.5)	70.6 (52.4–83.5)	<0.0001
BMI, (kg/m2)	27.7 (24.3, 31.0)	30.1 (26.7, 34.2)	0.09
Ex-smoker, n (%)	6 (35.3)	88 (91.7)	<0.0001
Ex-smoker pack years (n=94)	3.2 (0.5, 17.0) (n=6)	46.9 (33.0, 69.4) (n=88)	0.0002
ICS use, n (%)	-	88 (91.7)	-
ICS dose (Beclomethasone equiv. µg/day) (n=86)	-	2000 (1000, 2000) (n=86)	-
LABA use, n (%)	-	94 (97.9)	-
LAMA use, n (%)	-	89 (92.7)	-
SGRQ total	-	55.6 ± 17.4	-
CAT total	6.2 ± 4.7	21.1 ± 6.5	<0.0001
Pre β2 FEV1% predicted (n=112)	92.6 (89.4, 105.7) (n=17)	48.5 (39.1, 62.4) (n=95)	<0.0001
Pre β2 FVC % predicted (n=108)	91.8 (88.6, 103.2) (n=17)	74.4 (64.4, 83.2) (n=91)	<0.0001
Pre β2 FEV1/FVC % (n=108)	78.3 (72.8, 81.9) (n=17)	50.3 (36.3, 64.5) (n=91)	<0.0001
Post β2 FEV1% predicted (n=112)	95.8 (91.9, 106.9) (n=17)	51.8 (41.3, 66.5) (n=95)	<0.0001
Post β2 FVC % predicted (n=110)	96.2 (88.3, 104.0) (n=17)	77.6 (67.3, 86.7) (n=93)	<0.0001
Post β2 FEV1/FVC % (n=110)	82.0 (77.0, 85.5) (n=17)	52.5 (38.9, 65.0) (n=93)	<0.0001
Number of total exacerbations in past 12 months	-	2.0 (1.0, 3.5)	-
Number of severe exacerbations in past 12 months (n=95)	-	1.0 (0, 1.0)	-
GOLD quadrant A, n (%)	-	3 (3.1)	-
GOLD quadrant B, n (%)	-	22 (22.9)
GOLD quadrant C, n (%)	-	1 (1.0)
GOLD quadrant D, n (%)	-	70 (72.9)
Induced sputum analysis
Sputum cell viability, % (n=111)	68.9 (52.8, 80.4) (n=16)	80.6 (67.2, 92.8) (n=95)	0.01
Sputum total cell count, x 106/mL (n=111)	3.6 (1.7, 7.1) (n=16)	5.1 (2.9, 10.2) (n=95)	0.07
Sputum neutrophils %	25.4 (12.6, 34.9)	60.6 (42.1, 80.6)	<0.0001
Sputum eosinophils %	0.5 (0.1, 1.0)	1.9 (0.8, 4.0)	0.0001
Sputum macrophages %	65.1 (37.5, 75.0)	29.6 (15.8, 43.5)	<0.0001
Sputum lymphocytes %	3.0 (1.5, 5.4)	1.0 (0.3, 2.0)	0.001
Sputum columnar epithelial cell %	2.9 (0.4, 5.8)	2.0 (0.5, 4.5)	0.63
Peripheral blood eosinophil count (x109/L)	0.1 (0.1, 0.2)	0.2 (0.1, 0.4)	0.005

Notes: Data presented as n (%), mean ± SD or median (Q1, Q3). Bolding indicates significance (p-value <0.05).

Abbreviations: BMI, body mass index; ICS, inhaled corticosteroid; LABA, long-acting β2 agonist; LAMA, long-acting muscarinic antagonist; SGRQ, St George Respiratory Questionnaire; CAT, COPD Assessment Tool; FEV1, forced expiratory volume in one second; FVC, forced vital capacity.

MC/Basophil mRNA Abundance Increased Between COPD and Controls

Relative mRNA abundance significantly differed between COPD and control participants for CPA3, KIT, GATA2, SOCS2, ENO2 and HDC (Figure 1). CPA3, KIT, GATA2 and HDC mRNA abundance were increased in COPD vs controls. Control participants had increased SOCS2 and ENO2 mRNA abundance. No significant difference in the combinatorial gene metric was found.

Figure 1

Relative mRNA abundance of (A) TPSB2, (B) CPA3, (C) KIT, (D) GATA2, (E) SOCS2, (F) ENO2, (G) GPR56, (H) HDC, and (I) combinatorial MC/basophil gene metric in COPD participants compared to control participants. Bars represent median, with error bars representing Q1, Q3. Relative mRNA abundance compared between groups are expressed as 2−ΔCt relative to the housekeeping gene β-actin. Combinatorial gene metric based on ΔCt gene expression values. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001.

MC/Basophil mRNA Abundance Correlated in COPD

Spearman correlational analysis determined significantly positively correlated MC/basophil mRNA abundance between all genes in COPD sputum (Figure 2). Notable results include a strong positive correlation between TPSB2 and CPA3 (r=0.74, p<0.001), and moderate positive correlations between GATA2 and KIT (r=0.55, p<0.001), GATA2 and HDC (r=0.63, p<0.001) and HDC and KIT (r=0.69, p<0.001).

Figure 2

Correlational analysis heat map represents the correlations between MC/basophil-related gene mRNA abundance in COPD participants. Relative mRNA abundance 2−ΔCt values based on the housekeeping gene β-actin used in correlation analysis. Spearman correlation r-values reported. All correlations were significant.

Clinical Characteristics of Eosinophilic-COPD and Non-Eosinophilic COPD

E-COPD and NE-COPD were categorized using sputum differential cell count. Clinical characteristics of E-COPD and NE-COPD groups (Table 2) were not significantly different. Sputum neutrophil proportion was higher in NE-COPD vs E-COPD. Sputum eosinophil proportion and PBE count were higher in E-COPD vs NE-COPD.

Table 2

Clinical Characteristics of Eosinophilic COPD and Non-Eosinophilic COPD

	E-COPD	NE-COPD	p-value
Sample number (n)	44	52
Clinical characteristics
Sex, female n (%)	14 (31.8)	21 (40.4)	0.39
Age, years (range)	70.8 (52.4–83.5)	70.6 (54.4–82.2)	0.88
BMI, (kg/m2)	30.2 (27.3, 34.3)	29.1 (25.2, 34.2)	0.43
Ex-smoker, n (%)	39 (88.6)	49 (94.2)	0.46
Ex-smoker pack years (n=88)	42.0 (33.0, 62.5) (n=39)	53.8 (40.0, 75.0) (n=49)	0.05
ICS use, n (%)	40 (90.9)	48 (92.3)	1.00
ICS dose (Beclomethasone equiv. µg/day) (n=86)	1500 (1000, 2000) (n=39)	2000 (1000, 2000) (n=47)	0.27
LABA use, n (%)	44 (100.0)	50 (96.2)	0.50
LAMA use, n (%)	41 (93.2)	48 (92.3)	1.00
SGRQ total	54.7 ± 18.7	56.4 ± 16.4	0.63
CAT total	21.0 ± 6.9	21.2 ± 6.3	0.88
Pre β2 FEV1% predicted (n=95)	51.2 (41.9, 62.6) (n=43)	46.4 (33.2, 59.4) (n=52)	0.14
Pre β2 FVC % predicted (n=91)	74.7 (65.6, 82.9) (n=39)	74.4 (64.2, 85.9) (n=52)	0.95
Pre β2 FEV1/FVC % (n=91)	51.7 (41.4, 65.1) (n=39)	49.5 (31.3, 61.2) (n=52)	0.23
Post β2 FEV1% predicted (n=95)	54.2 (42.9, 67.4) (n=43)	50.5 (33.6, 62.0) (n=52)	0.21
Post β2 FVC % predicted (n=93)	79.6 (67.3, 86.5) (n=41)	76.7 (66.8, 87.0) (n=52)	0.93
Post β2 FEV1/FVC % (n=93)	55.6 (42.2, 66.4) (n=41)	52.0 (31.9, 61.6) (n=52)	0.12
Number of total exacerbations in past 12 months	2.0 (1.0, 4.0)	2.0 (1.0, 3.0)	0.31
Number of severe exacerbations in past 12 months (n=95)	0 (0, 1.0) (n=44)	1.0 (0, 1.0) (n=51)	0.05
GOLD quadrant A, n (%)	3 (6.8)	0 (0)	0.13
GOLD quadrant B, n (%)	8 (18.2)	14 (26.9)
GOLD quadrant C, n (%)	0 (0)	1 (1.9)
GOLD quadrant D, n (%)	33 (75.0)	37 (71.2)
Induced sputum analysis
Sputum cell viability, % (n=95)	79.2 (63.6, 87.7) (n=44)	86.7 (67.7, 94.5) (n=51)	0.06
Sputum total cell count, x 106/mL (n=95)	4.7 (2.8, 8.3) (n=44)	5.8 (3.0, 12.4) (n=51)	0.14
Sputum neutrophils %	53.2 (39.8, 63.8)	69.1 (47.1, 85.4)	0.0004
Sputum eosinophils %	4.9 (3.3, 13.3)	0.9 (0.5, 1.4)	<0.0001
Sputum macrophages %	32.9 (21.0, 42.9)	24.9 (10.4, 43.5)	0.11
Sputum lymphocytes %	1.3 (0.5, 2.1)	0.8 (0.3, 1.9)	0.33
Sputum columnar epithelial cell %	2.3 (1.0, 4.3)	1.3 (0, 5.1)	0.17
Peripheral blood eosinophil count (x109/L)	0.3 (0.2, 0.5)	0.2 (0.1, 0.2)	<0.0001

Notes: Data presented as n (%), mean ± SD or median (Q1, Q3). Bolding indicates significance (p-value <0.05).

Abbreviations: Eosinophilic COPD, E-COPD. non-eosinophilic COPD, NE-COPD. BMI, body mass index; ICS, inhaled corticosteroid; LABA, long-acting β2 agonist; LAMA, long-acting muscarinic antagonist; SGRQ, St George Respiratory Questionnaire; CAT, COPD Assessment Tool; FEV1, forced expiratory volume in one second; FVC, forced vital capacity.

MC/Basophil mRNA Abundance is Increased in Eosinophilic COPD

We compared MC/basophil gene expression between E-COPD, NE-COPD and control participants (Figure 3). Relative mRNA abundance differed for TPSB2, CPA3, KIT, GATA2, SOCS2, ENO2 and HDC between COPD inflammatory phenotypes and controls. GATA2 and KIT were significantly elevated in COPD inflammatory phenotypes compared to controls, however, showed no significant differences between the phenotypes. TPSB2, CPA3 and HDC mRNA abundance were significantly higher in E-COPD compared to NE-COPD and control participants. SOCS2 mRNA abundance was significantly increased in E-COPD compared to NE-COPD. SOCS2 and ENO2 mRNA abundance were significantly higher in control participants compared to NE-COPD. The combinatorial MC/basophil gene metric incorporating the 8 signature genes was significantly higher in E-COPD (−0.3070) vs NE-COPD (0.2141) vs controls (0.3139) (p=0.0002). Additional analyses comparing relative mRNA abundance of the genes between the four COPD airway inflammatory phenotypes (N-COPD [n=34], E-COPD [n=31], MG-COPD [n=13] and PG-COPD [n=18]) vs control participants [n=17] (, ) showed similar results. GATA2 and HDC were significantly elevated in disease inflammatory phenotypes compared to controls however, showed no significant differences between the phenotypes. For TPSB2 and CPA3, relative mRNA abundance was significantly higher in E-COPD compared to N-COPD, PG-COPD, and control participants. The combinatorial gene metric was significantly higher in the E-COPD (−0.4259) compared to the N-COPD (0.2809) and control participants (0.3139) (p=0.0006).

Figure 3

Relative mRNA abundance of (A) TPSB2, (B) CPA3, (C) KIT, (D) GATA2, (E) SOCS2, (F) ENO2, (G) GPR56, (H) HDC, and (I) combinatorial MC/basophil gene metric between Eosinophilic COPD (E-COPD), Non-eosinophilic COPD (NE-COPD) and control participants. Bars represent median, with error bars representing Q1, Q3. Relative mRNA abundance compared between groups are expressed as 2−ΔCt relative to the housekeeping gene β-actin. Combinatorial gene metric based on ΔCt gene expression values. #p<0.02 versus control participants; *p<0.02 versus NE-COPD participants.

MC/Basophil Gene Expression Predicts Airway Eosinophilic Airway Inflammation in COPD

Univariate logistic regression with ROC curve analysis was used to analyze the association of MC/basophil gene expression with eosinophilic vs non-eosinophilic airway inflammation (Table 3). All odds ratios produced were below 1, indicating increased gene ΔCt values (equating to decreased relative mRNA abundance) were associated with decreased odds of being categorized as eosinophilic. All genes (except KIT and ENO2) were significantly associated with eosinophilic inflammation. Univariate and multivariate LASSO regression determined CPA3 had the highest AUC of 0.78, suggesting it is the most important gene in predicting eosinophilic inflammation with good predictive accuracy. The combinatorial gene metric was also associated with eosinophilic inflammation (AUC=0.73, p<0.0001).

Table 3

Association Between Eosinophilic Inflammation Classification and MC/Basophil Gene Expression in COPD

	Univariate Logistic Regression		Multivariate Logistic Regression (LASSO)
Gene	Odds Ratio (95% CI), p-value	AUC	Coefficient	AUC
TPSB2	0.650 (0.519, 0.814), p<0.0001	0.77	0	-
CPA3	0.627 (0.499, 0.787), p<0.0001	0.78	−0.090	0.78
KIT	0.796 (0.616, 1.029), p=0.08	0.60	0	-
GATA2	0.759 (0.616, 0.936), p=0.01	0.63	0	-
SOCS2	0.702 (0.566, 0.870), p=0.001	0.73	0	-
ENO2	0.845 (0.681, 1.048), p=0.13	0.58	0	-
GPR56	0.758 (0.582, 0.987), p=0.04	0.61	0	-
HDC	0.673 (0.530, 0.854), p=0.001	0.70	0	-
Combinatorial gene metric	0.282 (0.140, 0.568), p<0.0001	0.73	-

Notes: Univariate logistic regression models of eosinophilic inflammation outcome and individual gene expressions (ΔCt relative to the housekeeping gene β-actin). Multivariate regression analyses performed with LASSO (least absolute shrinkage and selection operator). Combinatorial gene metric based on ΔCt gene expression values. Bolding indicates significance (p-value <0.05).

Abbreviation: AUC, area under the curve.

MC/Basophil mRNA Abundance is Increased in PBE-High COPD

We examined sputum MC/basophil gene abundance in relation to PBEs, using a clinically relevant cut-off (PBE high ≥300 cells/µL vs low <300 cells/µL). Clinical characteristics of PBE-high and low groups are shown in Table 4. For all genes, mRNA abundance was increased in PBE high (n=40) vs low (n=56) groups (Figure 4). The combinatorial gene metric was increased in PBE high (−0.3070) vs low (0.2004, p<0.0001).

Table 4

Clinical Characteristics of PBE-High and PBE-Low COPD

	PBE-High	PBE-Low	p-value
Sample number (n)	40	56
Clinical characteristics
Sex, female n (%)	15 (37.5)	20 (35.7)	0.86
Age, years (range)	70.0 (52.4–81.5)	71.4 (54.4–83.5)	0.06
BMI, (kg/m2)	30.4 (29.0, 35.1)	28.7 (25.2, 33.0)	0.14
Ex-smoker, n (%)	35 (87.5)	53 (94.6)	0.27
Ex-smoker pack years (n=88)	42.8 (22.0, 62.0) (n=35)	53.8 (40.0, 72.0) (n=53)	0.03
ICS use, n (%)	37 (92.5)	51 (91.1)	1.00
ICS dose (Beclomethasone equiv. µg/day) (n=86)	2000 (1000, 2000) (n=36)	1500 (500, 2000) (n=50)	0.01
LABA use, n (%)	40 (100.0)	54 (96.4)	0.51
LAMA use, n (%)	37 (92.5)	52 (92.9)	1.00
SGRQ total	58.0 ± 19.4	53.9 ± 15.8	0.26
CAT total	22.1 ± 7.2	20.3 ± 6.0	0.19
Pre β2 FEV1% predicted (n=95)	45.9 (38.8, 60.8) (n=39)	51.6 (39.8, 64.5) (n=56)	0.22
Pre β2 FVC % predicted (n=91)	73.1 (58.5, 82.4) (n=37)	76.1 (65.7, 87.1) (n=54)	0.14
Pre β2 FEV1/FVC % (n=91)	48.0 (38.8, 65.1) (n=37)	51.1 (35.2, 61.6) (n=54)	0.92
Post β2 FEV1% predicted (n=95)	49.0 (40.4, 61.5) (n=40)	53.2 (41.3, 67.9) (n=55)	0.30
Post β2 FVC % predicted (n=93)	74.9 (59.9, 84.9) (n=39)	78.6 (68.9, 87.4) (n=54)	0.15
Post β2 FEV1/FVC % (n=93)	52.7 (40.0, 65.3) (n=39)	52.4 (35.7, 62.7) (n=54)	0.66
Number of total exacerbations in past 12 months	3.0 (1.0, 5.5)	2.0 (1.0, 3.0)	0.19
Number of severe exacerbations in past 12 months (n=95)	0 (0, 1.0) (n=44)	1.0 (0, 1.0) (n=55)	0.62
GOLD quadrant A, n (%)	2 (5.0)	1 (1.8)	0.06
GOLD quadrant B, n (%)	5 (12.5)	17 (30.4)
GOLD quadrant C, n (%)	1 (2.5)	0 (0)
GOLD quadrant D, n (%)	32 (80.0)	38 (67.9)
Induced sputum analysis
Sputum cell viability, % (n=95)	80.6 (71.4, 94.1) (n=40)	80.8 (63.6, 92.0) (n=55)	0.67
Sputum total cell count, x 106/mL (n=95)	5.1 (3.1, 10.7) (n=40)	5.2 (2.7, 10.2) (n=55)	0.76
Sputum neutrophils %	58.1 (41.6, 80.6)	62.3 (42.4, 79.0)	0.66
Sputum eosinophils %	3.5 (1.9, 12.3)	1.0 (0.5, 2.3)	<0.0001
Sputum macrophages %	28.5 (15.8, 41.5)	33.1 (16.0, 45.8)	0.26
Sputum lymphocytes %	0.8 (1.3, 1.9)	1.3 (0.3, 2.1)	0.19
Sputum columnar epithelial cell %	2.0 (0.8, 3.9)	1.8 (0.3, 5.3)	0.96
Peripheral blood eosinophil count (x109/L)	0.4 (0.3, 0.6)	0.2 (0.1, 0.2)	<0.0001

Notes: Data presented as n (%), mean ± SD or median (Q1, Q3). Peripheral blood eosinophil (PBE) high (≥300 cells/µL). PBE low (<300 cells/µL). Bolding indicates significance (p-value <0.05).

Abbreviations: BMI, body mass index; ICS, inhaled corticosteroid; LABA, long-acting β2 agonist; LAMA, long-acting muscarinic antagonist; SGRQ, St George Respiratory Questionnaire; CAT, COPD Assessment Tool; FEV1, forced expiratory volume in one second; FVC, forced vital capacity.

Figure 4

Relative mRNA abundance of (A) TPSB2, (B) CPA3, (C) KIT, (D) GATA2, (E) SOCS2, (F) ENO2, (G) GPR56, (H) HDC, and (I) combinatorial gene metric in COPD participants categorized as PBE high or PBE low. Bars represent median, with error bars representing Q1, Q3. Relative mRNA abundance compared between groups are expressed as 2−ΔCt relative to the housekeeping gene β-actin. Combinatorial gene metric based on ΔCt gene expression values. PBE high (≥300 cells/µL) (n=40) and PBE low (<300 cells/µL) (n=56). *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001.

MC/Basophil mRNA Abundance is Associated with Lung Function

KIT, GATA2, GPR56, and HDC were negatively correlated with pre-β2 FEV1% predicted values, whilst GATA2 was negatively correlated with pre-β2 FEV1/FVC%, indicating higher relative mRNA abundance was correlated with lower lung function (Table 5). The combined gene metric was positively correlated with pre-β2 FEV1% predicted values, indicating a lower ΔCt value (indicating higher relative mRNA abundance) was correlated with lower lung function (Table 5). Similar correlations for KIT, GATA2, GPR56 and HDC were observed with post-β2 FEV1% predicted (data not shown). Increased GATA2 mRNA abundance was negatively correlated with BMI (r= −0.225, p=0.03). TPSB2 (r= −0.228, p=0.03) was negatively correlated with health status (CAT total score), indicating lower relative mRNA abundance was correlated with worse health status. MC/basophil mRNA abundance did not differ between COPD participants who experienced frequent (≥2) and infrequent (0–1) exacerbations, for total and severe exacerbations (Figure 5). The exception was GPR56, which was significantly increased in the group with infrequent severe exacerbations. The combinatorial gene metric was not significantly different for total or severe exacerbations.

Table 5

Associations Between MC/Basophil mRNA Abundance and Clinical Characteristics in COPD

	Age		BMI		SGRQ Total		CAT Total		Pre β2 FEV1% Predicted		Pre β2 FVC % Predicted		Pre β2 FEV1/FVC %
	r-value	p-value	r-value	p-value	r-value	p-value	r-value	p-value	r-value	p-value	r-value	p-value	r-value	p-value
TPSB2	0.079	0.44	−0.073	0.48	−0.136	0.19	−0.228	0.03	−0.035	0.74	0.008	0.94	−0.054	0.61
CPA3	0.009	0.93	−0.052	0.61	−0.122	0.24	−0.131	0.20	−0.003	0.98	0.008	0.94	−0.017	0.87
KIT	−0.025	0.81	−0.142	0.17	0.141	0.17	0.147	0.15	−0.283	0.006	−0.093	0.38	−0.199	0.06
GATA2	−0.190	0.06	−0.225	0.03	0.057	0.58	0.028	0.79	−0.315	0.002	−0.035	0.74	−0.241	0.02
SOCS2	−0.075	0.47	0.067	0.52	−0.047	0.65	−0.146	0.16	−0.091	0.38	−0.037	0.73	−0.020	0.85
ENO2	0.053	0.61	−0.008	0.94	0.091	0.38	0.120	0.24	−0.119	0.25	−0.175	0.10	0.022	0.84
GPR56	−0.076	0.46	−0.014	0.89	0.012	0.91	−0.064	0.53	−0.252	0.01	−0.054	0.61	−0.192	0.07
HDC	−0.010	0.92	−0.117	0.26	0.035	0.74	0.091	0.38	−0.220	0.03	−0.045	0.67	−0.181	0.09
Combinatorial gene metric	0.054	0.60	0.063	0.54	−0.037	0.72	0.011	0.91	0.227	0.03	0.074	0.49	0.151	0.15

Notes: Relative mRNA abundance 2−ΔCt values based on the housekeeping gene β-actin used in correlation analysis. Spearman correlation r-values reported. Combinatorial gene metric based on ΔCt gene expression values. Bolding indicates significance (p-value <0.05).

Abbreviations: BMI, body mass index; SGRQ, St George Respiratory Questionnaire; CAT, COPD Assessment Tool; FEV1, forced expiratory volume in one second; FVC, forced vital capacity.

Figure 5

Relative mRNA abundance of (A) TPSB2, (B) CPA3, (C) KIT, (D) GATA2, (E) SOCS2, (F) ENO2, (G) GPR56, (H) HDC, and (I) combinatorial gene metric in COPD participants classified by frequent (>2) and infrequent (0–1) exacerbator status for total (frequent [n=58] and infrequent [n=38]) or severe (frequent [n=14] and infrequent [n=81]) exacerbations. Bars represent median, with error bars representing Q1, Q3. Relative mRNA abundance compared between groups are expressed as 2−ΔCt relative to the housekeeping gene β-actin. Combinatorial gene metric based on ΔCt gene expression values. *p≤0.05.

Discussion

MCs and basophils have functionally diverse pathophysiological roles impacting on airway inflammation and remodeling. Their direct measurement in sputum is difficult due to their scarcity; hence, their inflammatory and clinical associations in COPD remain poorly understood. We measured a previously validated sputum MC/basophil gene signature in people with COPD and a control cohort and identified several MC/basophil-related genes increased in mRNA abundance in COPD participants. Correlation analysis revealed genes were all interrelated in COPD sputum. Increased MC/basophil gene expression was associated with airway and peripheral eosinophilic inflammation and decreased lung function in COPD. These novel findings demonstrate a dysregulation of MCs and basophils in E-COPD through measures of sputum MC/basophil gene expression.

Given their key roles in many airway pathological and inflammatory processes32–34 and their ability to activate via IgE-independent activation pathways,13,35 MCs and basophils may contribute to COPD pathogenesis. Our findings demonstrate dysregulation of sputum MCs and basophils in the airway lumen in COPD, and that sputum MCs and basophils are associated with both airway and peripheral eosinophilic inflammation in COPD. These findings are comparable to previous findings in severe asthma,14–17,36 demonstrating a common inflammatory mechanism involving MCs and basophils associated with eosinophilia in COPD and severe asthma. Notably, TPSB2, CPA3, HDC and the combinatorial gene metric were increased in E-COPD compared to NE-COPD. A recent study identifying SOCS2 and HDC as differentially expressed in E-COPD vs NE-COPD in sputum microarray analysis supports the positive association between MC/basophil genes and eosinophilic inflammation in COPD.37 This study also reported increased IL9R in E-COPD compared to non-eosinophilic COPD. IL9 promotes MC growth and differentiation, and this further supports the role of MCs and basophils in T2 inflammation in COPD.37 mRNA abundance of all MC/basophil genes was correlated in COPD sputum, supporting a common MC/basophil origin, with qPCR measures of MC/basophil-related genes reflecting MC and basophil abundance in sputum. There were no significant clinical differences between E-COPD and NE-COPD and therefore the differences in MC/basophil gene expression between the groups demonstrate the importance of molecular biomarkers such as these to distinguish between COPD inflammatory phenotypes. Previous asthma studies have shown networks of CPA3, HDC and GATA216 and TPSAB1 and CPA3,38 were related to T2 cytokine genes (IL-4, IL-5, IL-13) and future studies should address the potential role of MCs/basophils in promoting T2 inflammation in COPD. The latter study also demonstrated that IgE-independent epithelial IL-33 MC activation was related to T2 inflammation demonstrating potential importance of allergic and non-allergic MC-mediated mechanisms.38 We observed increased expression of all MC/basophil genes in PBE-high COPD. PBEs are promoted by IL-5 and IL-5 is expressed by MCs and basophils, further demonstrating a link between airway MC/basophils and a T2-high phenotype in COPD.39–41 CPA3 was found to be the best predictor of sputum eosinophilic inflammation in COPD. Therefore, the clinical measurement of MC/basophil signature genes in sputum is useful as a relatively non-invasive tool in predicting sputum inflammatory phenotypes in COPD. Surprisingly, SOCS2 and ENO2 expression were significantly increased in non-COPD controls compared to COPD participants. This may represent a mechanism of inflammatory dysregulation in COPD, as SOCS2 plays a role as a feedback inhibitor of IL1β and TLR signalling, particularly in dendritic cells42 and are also upregulated during dendritic cell maturation.43 Neither gene significantly correlated with age in COPD or non-COPD participants; thus, the lower age of the non-COPD controls likely does not fully account for this result.

KIT, GATA2, GPR56 and HDC were associated with decreased lung function. GATA2, KIT and HDC showed relatively high intercorrelation in COPD. GATA2, important in MC/basophil differentiation,44 promotes KIT45 and HDC46 expression, both important in MC/basophil development, proliferation and survival and acute inflammation.46,47 A previous study has reported FEV1% predicted values were decreased in eosinophilic COPD patients compared to non-eosinophilic COPD.8 As these genes were significantly increased in E-COPD, this may reflect the decreased lung function seen in E-COPD. Of note, TPSB2 and CPA3 whilst also increased in E-COPD and highly intercorrelated, were not associated with decreased lung function. Similarly, previous research has shown that CPA3 is not related to lung function in COPD.48 Airway MC expression of MC-related factors may be modulated in different inflammatory and clinical contexts in asthma,49,50 thus it may be that a particular MC/basophil phenotype associated with increased expression of GATA2, KIT and HDC is related to lower lung function in COPD, or that other factors that contribute to increased expression of these genes are also associated with lower lung function. GPR56 is a proposed basophil-specific gene.22 Its increased expression was related to poorer lung function and suggests a role of basophils in COPD. In contrast, GPR56 was the only MC/basophil gene differentially expressed between exacerbation phenotypes and was higher in participants with reduced severe exacerbation history. Previous findings have shown that basophil-derived IL-4, drives macrophage-produced MMP-12 and contributes to emphysema in COPD mouse models.34 Recent findings suggest a more complex scenario with a proposed immunomodulatory role for basophils in the airways.51 Lack of altered expression of other MC/basophil genes between exacerbation phenotypes suggests that MC/basophil presence in the airway lumen may not be an important predictor of exacerbation risk in COPD.

Our study has limitations. The cross-sectional study design does not provide an understanding of gene expression over time nor its longitudinal relationship to inflammatory measures. Replication of our findings in an independent cohort would be desirable in future studies. Retrospective data collection of exacerbations was based on patient recall over the past year. Atopic status and FeNO levels were not collected, and future work may provide further insight into the relationship of MCs and basophils to allergic COPD. Current smokers were excluded from our study, and validation of these findings in current smokers is important for future studies. The difference in age range between the control and COPD populations was large, with an older median age in COPD participants. We do not know the influence of age on MC/basophil gene expression. However, we found no correlation between MC/basophil mRNA abundance and age in COPD participants, nor did age differ significantly between E-COPD and NE-COPD.

Conclusions

In conclusion, we applied a sputum qPCR-based MC/basophil gene signature validated in asthma, to a COPD cohort compared to controls and demonstrated that sputum MCs and basophils were dysregulated in COPD, with several genes increased in COPD participants. We also found that MC/basophil gene expression was associated with airway and peripheral eosinophilic inflammation and poorer lung function in COPD. The clinical measurement of MC/basophil signature genes in sputum is useful as a relatively non-invasive tool for endotyping E-COPD, with clinical associations with decreased lung function.